The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.
Human vision is readily capable of discerning between structures and surrounding areas, even in blurred video images due to elements such as fog and dust. However, images captured by a video capturing device in such an environment tend to lose most of feature points. The inventor(s) has noted that to reduce the gap that exists between the image and the human visual perception capability, there have been various studies conducted, proposing Retinex algorithms as a result. The inventor(s) has noted that mimicking the color discrimination mechanism of the human visual system, the Retinex algorithm is a method for separating the human influential illumination component from a reflection component which reflects the characteristics of the image for removing the illumination component effect and emphasizing the reflection component to compress the dynamic range of the input image and improve the contrast.
Meanwhile, the Retinex algorithm estimates the illumination component of the image by using a Gaussian filter. The inventor(s) has noted that when the input image is in color, it is necessary to repeatedly apply each of the color components R, G and B to the Gaussian filter. Furthermore, the inventor(s) has noted that when utilizing a plurality of Gaussian filters with different filter coefficients, each of the Gaussian filters needs to process each of the color components R, G and B, which increases the computation amount, thereby preventing image correction from being processed in real time. The inventor(s) has experienced that the higher the resolution of the image gets, the more ineffective the method becomes.